netdev - Re: [PATCHv2] tls: Add support for encryption using async offload accelerator

lists.openwall.net		lists / announce owl-users owl-dev john-users john-dev passwdqc-users yescrypt popa3d-users / oss-security kernel-hardening musl sabotage tlsify passwords / crypt-dev xvendor / Bugtraq Full-Disclosure linux-kernel linux-netdev linux-ext4 linux-hardening linux-cve-announce PHC
Open Source and information security mailing list archives

Hash Suite: Windows password security audit tool. GUI, reports in PDF.

[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]

Date:   Wed, 31 Jan 2018 09:34:44 -0800
From:   Dave Watson <davejwatson@...com>
To:     Vakul Garg <vakul.garg@....com>
CC:     "linux-crypto@...r.kernel.org" <linux-crypto@...r.kernel.org>,
        "ilyal@...lanox.com" <ilyal@...lanox.com>,
        "aviadye@...lanox.com" <aviadye@...lanox.com>,
        "davem@...emloft.net" <davem@...emloft.net>,
        "netdev@...r.kernel.org" <netdev@...r.kernel.org>,
        Gilad Ben-Yossef <gilad@...yossef.com>
Subject: Re: [PATCHv2] tls: Add support for encryption using async offload
 accelerator

On 01/31/18 05:22 PM, Vakul Garg wrote:
> > > On second though in stable we should probably just disable async tfm
> > > allocations.
> > > It's simpler. But this approach is still good for -next
> > >
> > >
> > > Gilad
> > 
> > I agree with Gilad, just disable async for now.
> > 
> 
> How to do it? Can you help with the api name?

*aead = crypto_alloc_aead("gcm(aes)", 0, CRYPTO_ALG_ASYNC);

https://github.com/ktls/net_next_ktls/commit/f3b9b402e755e4b0623fa83f88137173fc249f2d

> > If the flag MSG_DONTWAIT is set, we should be returning -EINPROGRESS and
> > not wait for a response.  I had started working on a patch for that, but it's
> > pretty tricky to get right.
> 
> Can you point me to your WIP code branch for this?

https://github.com/ktls/net_next_ktls/commit/9cc839aa551ed972d148ecebf353b25ee93543b9

> If MSG_DONTWAIT is not used, will it be sane if enqueue the crypto request to 
> accelerator and return to user space back so that user space can send more plaintext data while 
> crypto accelerator is working in parallel?

Right, that's roughly what the above does.  I believe the tricky
unfinished part was getting poll() to work correctly if there is an
async crypto request outstanding.  Currently the tls poll() just
relies on backpressure from do_tcp_sendpages.